Menu

The Digital Fairness Act: What the EU’s Consumer Rulebook Means for AI

The Digital Fairness Act: What the EU’s Consumer Rulebook Means for AI

4 min read

|

November 19 2026

4 min read

|

November 19 2026

The Digital Fairness Act: What the EU’s Consumer Rulebook Means for AI

On 11 November 2026, the European Commission proposed the Digital Fairness Act (DFA). It is a regulation that would repeal and replace the 2005 Unfair Commercial Practices Directive. Most coverage has focused on dark patterns, loot boxes and influencer disclosures. For AI policy, the more interesting story is how the proposal treats automated systems, chatbots and personalisation.

 A rulebook for the attention economy

The Commission’s starting point is that the 2005 framework predates the data and attention economy, influencer marketing, personalisation and subscription-based digital services. The 2024 Fitness Check found that consumer law remains relevant but only partially effective, with complaints and detriment still high.

Moving from a directive to a regulation matters. One directly applicable text would replace 27 national transposition regimes. It would also create the legal precondition for the Commission to enforce directly in specific cases under the proposed revision of the Consumer Protection Cooperation Regulation.

Where AI shows up

Agents that buy on your behalf. Article 26 is perhaps the most forward-looking provision. Where a trader offers consumers an automated system that concludes and performs contracts on their behalf, the consumer must expressly choose on first use between confirming every contract, confirming only certain types (with criteria such as a maximum price and a time limit), or no confirmation at all. The consumer can change these parameters at any time and receives confirmation of the choice on a durable medium. The system must also disclose any payment the trader receives for proposing a contract, and any conflict of interest. This is a consumer-law answer to the agentic commerce question: who is in control when software shops for you?

Addictive Design. Article 17 would require traders whose products encourage compulsive or excessive use to give consumers tools to switch off specific features. One of them is behaviour that simulates interpersonal relations, such as persistent prompts to keep engaging with an automated system. The recitals say AI companions and general conversational chatbots may simulate interpersonal relationships to incentivise prolonged use. For minors, the proposed EU KIDS Act would take the lead on such services, so the DFA mainly extends the switch-off right to adults.

A human on request. Article 25 says that where a trader provides online customer service after a contract is concluded, a human must be reachable on request when other means are not effective. Anyone deploying support chatbots should note it.

Profiling and personalisation. Article 12 would prohibit advertising targeted at minors based on profiling, and advertising based on special categories of personal data where it could harm the consumer. It extends to other traders prohibitions that the Digital Services Act already imposes on platforms. Article 13 prohibits personalised pricing based on profiling when it could harm minors or vulnerable consumers. The proposal acknowledges that personalisation can correlate with protected characteristics, and responds through transparency and a ban on exploiting vulnerability.

Dark patterns and the burden of proof

Article 10 prohibits designing interfaces in a way that is likely to deceive or manipulate consumers, or otherwise materially distort their economic behaviour. A list of practices banned in all circumstances sits in Annex I, and the Commission could update it by delegated act as technology and market practice evolve. For AI-driven interfaces that adapt in real time, a flexible, updatable list is arguably the only workable approach.

The Commission’s evidence base helps explain the urgency. A 2022 Commission study found that 97% of the most popular websites and apps used by EU consumers deployed at least one dark pattern.

Article 34 may prove just as consequential for algorithmic systems. Where a claimant presents facts and evidence sufficient to support the likelihood of an unfair practice online, or an authority has reasonable suspicion, courts and authorities can require the trader to disclose evidence on the design, operation and impact on consumer behaviour of the practice. Courts may also treat a failure to disclose as relevant when assessing unfairness, subject to trade secret protections. Personalised systems are hard for outsiders to audit, and this is the proposal’s attempt to rebalance that. The memorandum also describes a general “fairness by design” obligation, not tied to specific process requirements.

 Enforcement and stakes

For widespread infringements, national maximum fines would be set at no less than 4% of annual turnover in the Member States concerned. The Commission could fine up to 6% of worldwide annual turnover where it investigates under the revised CPC Regulation. Enforcement of Article 17(1) would sit with Digital Services Act authorities for online platforms, and with AI Act authorities for providers and deployers of AI systems in scope.

The Commission estimates a reduction in consumer detriment of EUR 88.4 billion over 2025-2035, or EUR 13.7 billion a year including monetised time losses, against recurrent business costs of EUR 1.2 billion a year (EUR 678 million adjusted for overlaps). Press coverage notes that business groups, including games and gambling interests, have criticised the proposal.

What to watch

The DFA is a proposal, not law. If adopted, it would apply 12 months after entry into force. Several questions are worth following in Parliament and Council:

  • Overlap. How will the DFA fit with the AI Act, the DSA and the EU KIDS Act, and will the without duplicating promise hold in practice?

  • Agent rules. Will Article 26 survive negotiations, and how will it work with fast-moving agent products?

  • Evidence disclosure. How will Article 34 be balanced against trade secrets?

  • Vague triggers. How will “can reasonably be foreseen to encourage compulsive or excessive use” be interpreted?

For AI developers, the practical message is that consumer law will reach conversational and agentic products through design choices, defaults and user controls, not only through the AI Act’s risk categories.


For keen followers of AI4POL, WP2 is working on a task on the legal and ethical assessment of online consumer manipulation under the AI Act and consumer law. The DFA sits at the epicenter of this question. The study is currently in progress and we will share results on this as of next June. 

by

Pratiksha Ashok

/

This project has received funding from the European Union’s Horizon Europe research and innovation program under grant agreement No 101177455

This project has received funding from the European Union’s Horizon Europe research and innovation program under grant agreement No 101177455

This project has received funding from the European Union’s Horizon Europe research and innovation program under grant agreement No 101177455